Developing a quantitative framework for implementing traffic management initiatives in the National Airspace System

Abstract

The National Airspace System (NAS) constitutes a complex system comprising dynamic parameters, multiple stakeholders and sources of uncertainty. Varying weather and traffic conditions often cause unexpected system-wide delays and disruptions, which can lead to significant costs for airlines and passengers. Air traffic managers facilitate the balance of demand and capacity within this complex system by continuously monitoring numerous sources of information that change in real time, and by constantly communicating and coordinating with a myriad of stakeholders, such as facilities and airline dispatchers, through the process of Collaborative Decision Making (CDM). During periods when demand exceeds capacity, air traffic managers have a toolbox of strategies called traffic management initiatives (TMIs) that can be utilized to manage elements of the NAS. This toolbox includes ground delay programs (GDP), ground stops (GS), airspace flow programs (AFP) and strategic reroutes. By assessing multiple parameters and sources of uncertainty simultaneously, traffic managers consider their options and alternatives at several decision points in order to choose which of these numerous strategies to implement. Currently traffic managers largely rely on operational experience and tacit knowledge when deciding which TMIs to implement. However, as the NAS progresses towards NextGen goals such as Collaborative Airspace Constraint Resolution (Cactus) and Collaborative Air Traffic Management Technologies (CATMT), it will be increasingly important to ensure that CDM processes, including TMI decision-making, are systematic and repeatable. Here we conduct a holistic decision analysis (DA) of TMIs to systematically identify the elements of air traffic management decisions, and to provide a rigorous assessment of possible outcomes and sources of uncertainty. An influence diagram illustrates a generalized view of the decision network and identifies the decision elements, including sources of uncertainty and consequences. The influence diagram is then translated into Bayesian networks to probabilistically model observed relationships and conditional dependencies, and evaluate the likelihood of the chance elements. Ultimately, the Bayesian network is integrated into probabilistic decision trees representing each element at specific decision points in time. Importantly, the decision points span several hours prior to an event ranging from the strategic to tactical environments to account for varying uncertainty levels over time, and to incorporate the balance between efficiency and predictability. By identifying and quantifying uncertainty in TMI decision-making, developing a decision framework, and characterizing potential outcomes, we aid the decision-making process for air traffic managers and provide a resource for more methodical TMI implementation, coordination with stakeholders, and congestion management in response to variable weather and demand conditions.